I'm not sure if this link has been posted before, but it may be of interest to some:
The General Feedback Theorem: A final solution for feedback systems, by Middlebrook
https://core.ac.uk/download/pdf/4872885.pdf
The General Feedback Theorem: A final solution for feedback systems, by Middlebrook
https://core.ac.uk/download/pdf/4872885.pdf
Don't be disheartened if you can't make sense of the maths.
I am now reasonably convinced that some of it doesn't actually make sense.
Certainly I haven't found a way to interpret it that makes sense and no-one else has chimed in despite some clever people in the thread, we shall see.
I do appreciate the prod to review my old maths/physics but I do find it an effort so thanks for the support.
Best wishes
David
Hi Ian
That one is worth another mention even if it has been posted previously.
Nice to see you because this thread is essentially a continuation of my attempts to understand some of the "weird" phase plots that we both noticed earlier.
I have made some advances, still not totally clear but I believe the "null loop" is the last piece of the puzzle.
Best wishes
David
If I could make a comment, sometimes better includes behavior under sudden changes in large signal non-linearities. Any of these feedback techniques could, for instance, oscillate at 5V from one or the other rail under load and behave fine in between.
David, have you analyzed .asc of my tt amp you ask for? It useses a bit different compensation, TPC around enhanced VAS plus touch of Cherry from the output to the common TPC point, I called it OITPC (Output Inclusive Two Pole Compensation). The same compensation I use i my working CFA and it works like a charm.
BR Damir
Yes, For specific circuit simulations I do Bode plots of closed loop Return Ratio as I push the output from rail to rail (more or less) to check stability under exactly these conditions.
But the theory is essentially linear, I don't see that Miller compensation, for instance, will be different to Input Inclusive in this respect.
But thanks for the point, I will think about it.
By the way, didn't you study Physics at MIT before your career at AD, does Waly's explanation look correct to you?
The superficially plausible phrases don't convince me.
Best wishes
David
Last edited:
Your compensation is one of the principle reasons I need to work out null loop theory
All the interconnected loops are not possible to analyse with simple loop theory.
I will try to use it as a test case.
Best wishes
David
No I'm afraid some reading would be in order. BTW it is true and we have seen it that, say, an unstable bias spreader inside a loop will not have its resonance excited in normal simulations. I also disagree if comparing something like output inclusive plus Vas vs. Vas only if the problem is output device quasi-sat the two loops near the rail could look very different than at 0 output.
How do you check for this so that it doesn't surprise you in a physical circuit?
Do you have some "abnormal" simulation?
Yes, the example I had in mind was a common emitter output.
The simple linear feedback model does not handle this AFAIK.
That's why I do the simulator checks.
Best wishes
David
One possible explanation would be that some of your reply doesn't actually make sense.
Then you would not be surprised that I can't make sense of it.
Now if you answer my question from post #78, repeated from post #60, we can move forward.
Best wishes
David
I think there's a much simpler explanation: you have no real interest in understanding the subject, other than stirring the pot.
You know very well the answer to your own question. Me, I'm no longer playing, and I'm sorry I dared again to pollute your thread with my nonsense (actually coming straight from books you don't care to open). BTW, audio amplifiers are still, and always, minimum phase systems.
All a modeling issue, this was an old planar process where small pockets of super-integration were common. The bias spreader had very little phase margin in its own tight loop but because the silicon was modeled as separate devices it did not oscillate in the simulator but added only an insignificant bump outside of the unity gain frequency and was ignored. If it had all out oscillated it would have shown up.
On the contrary, I will even pay you, or anyone else, for a lucid explanation of your claims.
I will still pay even if the explanation reveals that your position is incorrect but makes sense.
I am happy to spend the price of a textbook for this sort of education, say $50, half if incorrect.
No, and it would appear from the lack of comments, nor does anyone else.
$10 for an answer to this bonus question
No trouble, it makes me think about the problem and the posts provide a useful way to record my ideas.
So, to write while it's in my mind.-
I think you have conflated two ideas of "volume"
One use of "volume" is as a set of points, say a spherical volume of space, all the points within one radius of the center.
The other use is the measure of that volume, 4/3 Pi radius^3.
Your phase space comments seem to refer to a volume in the first sense.
But the determinant is a measure, that's why I say I can't make sense of your posts.
I am keen to read any reference you can provide, and well timed I say, I'm off to the library soon.
What do you recommend?
Best wishes
David
Last edited:
Just passing by, your understanding of "volume" in the context of phase space makes me think you don't need a book, you need about two years of post graduate math. Hint: a "volume" is a differentiable form of the same "grade" as the phase space dimension that is #0 everywhere.
Space phase has, due to it's oriented Riemannian properties, a natural volume V=SQRT(|D|)dX1^dX2^...^dXn where dXi are in general 1-forms forming a basis (which are in this particular case resulting by applying the phase space method for solving the linear differential equation, by rewriting the original equation as a system of differential equations that are first order in time, by introducing additional variables, etc...) and |D| the absolute value of the determinant of the matrix representation of the metric tensor (which is in this particular case the matrix associated with the linear differential operator).
Go ahead and nitpick about the poor notations, lack of defining 1-forms and metric tensor. Now that I'm thinking there is likely a more intuitive path:
Circuit->signal graph->matrix->matrix characteristic polynomial (roots are the eigenvalues)->differential equation (simply substituting p->d\dt)->space phase method for solving the differential equation->volume in the space phase using the matrix determinant and the 1-forms in the linear equation set.
BTW, qualifying things you don't understand (which given the nature of the beast, is not a crime) as "doesn't actually make sense" is impolite if not straight rude (not to mention the "award" for whoever proves your discussion peer wrong). If you had any questions rooting in your lack of experience and/or education in these issues, a common polite method is a private message or email. And it's not the first time you have the same approach.
And now I'm really over and out. This discussion has absolutely no sense on this forum.
You misunderstand, it's a reward for whoever corrects my mistakes or improves my education.
Payment is for an explanation of your point, proof my suspicions were ill founded.
The money is less if they prove you incorrect
I pay for textbooks, seems only fair to repay anyone, you included, who takes the time and effort to explain a point about which I am unclear.
Best wishes
David
The 1-form stuff has sent me back to R. Penrose, which I had not checked.
Perhaps I need your PayPal account.
Last edited:
Back to feedback.
Here is an interesting article about multiple-loop feedback and a LTspice
implementation:
Multiple-Loop Feedback Analysis | Electronics Insights | Rob Fox
Best wishes Josef
Here is an interesting article about multiple-loop feedback and a LTspice
implementation:
Multiple-Loop Feedback Analysis | Electronics Insights | Rob Fox
Best wishes Josef
Indeed
Thank you, as always, for the article.
It is a new one to me, I shall read it with interest.
Best wishes
David
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.